How India Built One AI Network to Run Its Russian, Western, and Homegrown Air Defenses

The early-morning calls came into the Indian Army's short-range air defence crews on the western border last May, not as voice traffic but as machine-paired assignments. Software, not gun crews, was deciding which surface-to-air battery would take which incoming target. The system in the middle of that loop is Akashteer, an AI-enabled command and reporting layer that fuses Soviet-era, Western, and Indian-made radars and weapons into one shared picture of the sky and pairs each threat with the best-suited weapon in the time it takes a sensor to sweep the horizon.

That was the operational test for a system the Indian Ministry of Defence had been buying in stages. The MoD had signed two contracts worth a combined Rs 2,400 crore (about $285 million at recent 2025 exchange rates, approximate) with state-owned integrator Bharat Electronics Limited (BEL) for the Automated Air Defence Control and Reporting System, known internally as Project Akashteer, according to a PIB primary press release. During Operation Sindoor, the cross-border engagement on India's western border in 2025, the system coordinated the short-range air defence units that engaged incoming fire. BEL's official corporate communications said in a post on X that Akashteer made life "hell" for Pakistani air adventures and that ground-based air defence systems integrated with Akashteer "proved their mettle in the war-field."

Akashteer is, in mechanical terms, a software problem masquerading as a hardware story. The Indian Army does not buy its air defence from one country. Its radar inventory mixes Russian-origin systems, Western acquisitions, and indigenous DRDO-developed units. Each has its own data formats, refresh rates, identification friend-or-foe keys, and engagement procedures. Without a layer that translates all of those into one common operational picture, an operator looking at a screen has to mentally stitch together what the P-18 radar sees, what the indigenous 3D surveillance radar sees, and what the weapon battery's own tracker sees, all while the threat is closing. Akashteer does that stitching, then hands the matched threat-weapon pair to the right crew with the engagement geometry already computed. DRDO and the Indian Space Research Organisation (ISRO) co-developed the data-fusion and communications backbone, while BEL handles production and integration.

The reaction-time figure that has travelled the furthest, two to three milliseconds from threat detection to weapon assignment, comes from IDRW reporting and is best read as an order-of-magnitude claim rather than a verified engineering benchmark. In context, the 2–3 ms figure describes the software's internal pairing decision after radar and identification data arrive, not the end-to-end kill chain, which still waits on human authorisation, weapon slew time, and flight time to target. Primary engineering documentation for the latency number has not been published. Even with that caveat, the operational difference is the point: the system is fast enough that the human is no longer the bottleneck for picking which launcher takes which target, only for the final engagement decision. DRDO's chief has publicly floated international demand for Akashteer after the Sindoor performance, a notable signal from a lab that usually frames export potential more cautiously.

The system does one job that does not photograph well but matters in any combined-arms fight. By holding a single, time-stamped picture of every friendly radar return and every weapon battery's engagement radius, Akashteer is built to keep one Indian unit from firing on another. Fratricide, friendly fire between own units, has been a documented risk in every modern air defence network, and the credible claim for Akashteer is that it gives every connected control room the same picture of where every friendly return sits. The Indian Army is now in phased induction of the system, replacing the older, manually-coordinated reporting chain.

Three limits are worth carrying into any reading of the Sindoor performance. First, one operational episode is not a stress test; air defence networks are typically validated across contested electronic-warfare environments, saturated raid sizes, and multi-axis attacks, and the public reporting on Sindoor does not give a clear view of which of those Akashteer actually faced. Second, the system has a single integrator. BEL builds it, DRDO and ISRO co-developed the layers beneath it, and the Indian Army is the customer. That concentration is a procurement success and a supply-chain risk at the same time. Third, the most enthusiastic coverage of Akashteer, including much of the framing in the Indian press, comes from outlets with a pro-Indian editorial line, Organiser being the clearest case, and the engineering benchmark for the headline latency number has not been independently published.

The next data point will be the first export customer, if one signs, and the first independently published benchmark for the system's pairing latency under load. Until then, Akashteer's case rests on one short, intense, and under-documented engagement on India's western border, and on the unglamorous claim that the crews who used it did not shoot at each other.